Voice notes that write themselves

But people don't always want everything. When we speak, we repeat ourselves. We pause. We change our minds. We use filler words. A transcript keeps all of it.

So the user is left with another job.

The recording was easy. The useful part came later, and I wanted to shorten that gap.

What if you could speak naturally and get something useful back?

The basic idea is five steps.

The user does not need to think about the steps in the middle. They just speak. Handy does the heavy lifting, and they get a result they can review, change and use.

The more I thought about it, the clearer that became. There are already many tools that turn speech into text. The bigger problem is what happens after that, because a transcript is not the same thing as a useful note.

Someone might speak for five minutes about a project. Inside that recording could be three ideas, two tasks, a problem they noticed, and a decision they need to make. All of that information is there. It just isn't organised.

So Handy should not stop at transcription. It should help turn the raw thought into something easier to use.

That meant Handy had to do three things well.

That became the foundation for the product.

You open Handy. You see one main action. Talk.

There was no need to explain the product before someone used it. The user already knows how to talk, so the interface gets out of the way.

A recording could look like this:

The idea was clear. The sentence was not. Handy cleans up the unnecessary parts while keeping the original meaning.

Clean the thought, not the thinking.

A person might talk about several things in one recording, so Handy finds the different parts of the thought and separates them.

The goal was never to make the user start over. It was to give them a better starting point.

Cleaning a sentence is only half of it. How clean it should sound depends on where it is going.

If you are dictating into Gmail, the thought should come out as something you could send to a colleague. If you are in a chat with a friend, it should not. It is the same recording, dressed for a different room.

You do not copy anything. You do not switch windows. You speak where you were already working, and the words turn up in the register that place expects.

In settings it is one switch, Context-Aware Cleanup, sitting under a style you can set by hand. Reading the room is the default, not the only option: if you would rather decide for yourself, you pick the style and Handy stops guessing.

The tool should fit the room you are in, not ask you to come to it.

AI does a lot of the work in Handy. But the result is a draft, not a verdict. Someone can review it, edit it, remove something, add something, and then use it wherever they need it.

That made the product feel less like an AI that writes for you and more like a tool that helps you get your thoughts into shape.

I used Claude Code to turn the designs into a working product, which meant I could move quickly between design and code.

Something could look good in Figma and feel wrong the moment I used it. So I changed it. I could also test ideas much earlier, instead of waiting for a developer to build them, which made the design work more hands-on than it had ever been.

Handy started with a simple problem: speaking is fast, but turning speech into useful work takes time.

You speak. Handy turns your words into clean, structured content. You review it, you make it yours, and then you use it.

The product is simple on the surface. Underneath it is a full process for moving from a rough thought to something useful.

It just needs to remove the parts that get in the way. For Handy, that meant taking care of the messy work between speaking and having something useful.

The user still owns the thought. Handy just helps them get there faster.